K. M. Sunjiv Soyjaudah

Effects of prior learning on mathematics performance at secondary level

This research study was triggered by the relatively low percentage of students who successfully study mathematics at the Higher School Certificate (HSC) level. While pass rate in mathematics averages 76% at primary level, it remains a concern that only around 42% of School Certificate (SC) students get a good grade (distinction or credit) in that subject. Causes are multifaceted but this study particularly sought to investigate the effects of prior learning achievement at the Certificate of Primary Education (CPE) level on academic performance at secondary level in Mathematics. It analyzed the distribution of grades at SC and established a relationship with results obtained five years before by the same group of students. The trend was analyzed using data from the Mauritius Examinations Syndicate (MES) for three cohorts of CPE students. Compared to the number of students taking the CPE exams, it was found that only 18% of them studied the subject at HSC level. Grades at SC proved to be significantly influenced by CPE grades. Students consistently regressed in their mathematics learning over the five years of secondary schooling. This study unveils that to some extent mathematics disabilities can be diagnosed right at CPE level and confirms that prior knowledge has a determining effect on future learning achievement.

Self Monitoring Analysis and Reporting Technology (SMART) Copyback

Users have variety of requirements of data storage that can be addressed using Self Monitoring Analysis and Reporting Technology (SMART) Copyback. It is estimated that over 94% of all new information produced in the world is being stored on magnetic media, most of it on Physical Disks (PD). Moreover, larger population studies rarely have the infrastructure in place to collect health signals from components in operation, which is critical information for detailed failure analysis. It presents the data collected from detailed observations of a large disk drive population in production Internet services deployment. Analysis identifies several parameters from the Physical Disks (PD), self monitoring facility (SMART) that correlate highly with failures. Despite this high correlation conclude that models based on SMART parameters alone are unlikely to be useful for predicting individual drive failures. Surprisingly, it found that temperature and activity levels were much less correlated with Physical Disk (PD) failures.

Enhancing the error-correcting performance of LDPC codes for LTE and WiFi

Low-density parity-check (LDPC) codes are one of the most powerful error-correcting codes for wireless communication of high-speed data. This power comes from an efficient decoding scheme, the sum-product algorithm (SPA), which runs over the bipartite graph defining the code. In this work, we propose to use a modified version of the decoding algorithm over the LDPC codes used in Long Term Evolution (LTE) and WiFi and demonstrate that with the same number of iterations, without any increase in time complexity, we can achieve an enhanced error-correcting performance. The only trade off of the system is the necessary storage of the decoding sets, for a total size of ns × n, where ns is the number of decoding sets used and n is the codeword length. In this work, the maximum value of ns used is 5, which makes the trade-off fairly acceptable, considering that the coding gain is obtained virtually free-of-charge.

A VPN framework through multi-layer tunnels based on OpenSSH

This paper details how to setup and test the new tunneling features of OpenSSH to establish an enhanced SSH Layer 3 VPN between three computers in a Wide Area Network (WAN) environment. The OpenSSH security feature will be explored to provide secure tunneling and different authentication methods. Using OpenSSH to built VPN will cater for security by encrypting the data transmitted across the public network to private networks. OpenSSH VPN is said to have limitations with numerous clients and therefore it is advisable to use OpenVPN when dealing with more than one clients. To the best knowledge of the authors, this is the first reported successful implementation of a wide-area network VPN using OpenSSH tunnels with multiple clients. This proof-of-concept positions OpenSSH therefore as a potential alternative to the more conventional OpenVPN.

Application of Predictive Coding in Neuroevolution

This paper presents promising results achieved by applying a new coding scheme based on predictive coding to neuroevolution. The technique proposed exploits the ability of a bit, which contains sufficient information, to represent its neighboring bits. In this way, a single bit represents not only its own information, but also that of its neighborhood. Moreover, whenever there is a change in bit representation, it is determined by a threshold value that determine the point at which the change in information is significant. The main contributions of this work are the following: (i) the ratio of the number of bits to the amount of information content is reduced; (ii) the complexity of the overall system is reduced as there is lesser amount of bit to process; (iii) Finally, we successfully apply the coding scheme to NEAT, which is used as a biometric classifier for the authentication of keystroke dynamics

A Hybrid Genetic Algorithm and Radial Basis Function NEAT

We propose a new neuroevolution technique that makes use of genetic algorithm to improve the task provided to a Radial Basis Function – NEAT algorithm. Normally, Radial Basis Function works best when the input- output mapping is smooth, that is, the dimensionality is high. However, if the input changes abruptly, for example, for fractured problems, efficient mapping cannot happen. Thus, the algorithm cannot solve such problems effectively. We make use of genetic algorithm to emulate the smoothing parameter in the Radial Basis function. In the proposed algorithm, the input- output mapping is done in a more efficient manner due to the ability of genetic algorithm to approximate almost any function. The technique has been successfully applied in the non- Markovian double pole balancing without velocity and the car racing strategy. It is shown that the proposed technique significantly outperforms classical neuroevolution techniques in both of the above benchmark problems.

Readiness of mathematics' educators to incorporate ICT as a teaching tool in Mauritius

Information and Communications Technology (ICT) is opening avenues for disseminating content in innovative and effective manners to learners. The overarching construct namely technology, pedagogy and the digital content are key elements for the successful integration of ICT in school settings. In this paper we investigate pedagogy and more precisely the readiness of educators to adopt technology in their teaching. A questionnaire was administered to 118 mathematics educators working in 95 secondary schools. The results reveal that mathematics educators in Mauritius are regular users of computers at home and are much willing to integrate this tool in their teaching process if a conducive environment is provided in the schools, including facilities, training and motivation. Furthermore educators strongly perceived ICT affordances for learning benefits and believed that students can learn mathematics better with ICT tools.

Parallel concatenation of recursive systematic convolutional codes with transmission of interleaved set

In this work, a scheme for the parallel concatenation of convolutional codes is proposed. The systematic and interleaved sets are encoded. Unlike conventional turbo codes, however, both encoded sequences are concatenated and transmitted. At the receiving side, the first half of the corrupted sequence (corresponding to its encoded counterpart at the transmission side) is used as input to a soft-in soft-out (SISO) decoder. The second half is, in a similar manner, used by a second SISO decoder. These two decoders then use the maximum a-posteriori (MAP) algorithm for decoding. The key difference, however, with conventional turbo schemes, is that the channel metric is not removed when computing the extrinsic information for the first iteration. This is made possible since both decoders have access to an independently corrupted set. In the sequel, there is almost no gap in the performance with increased signal to noise ratio. In addition, for the remaining iterations, the proposed scheme employs conventional decoding and makes use of an inherent stopping criterion that enables the decoding procedure to be stopped once both decoders converge to the same sequence. Thus, the system makes an efficient use of iterations while reducing the computing resources required for decoding.

How to improve performance of Neural Network in the hardened password mechanism

A wide variety of systems, ubiquitous in our daily activities, require personal identification schemes that verify the identity of individual requesting their services. A non exhaustive list of such application includes secure access to buildings, computer systems, cellular phones, ATMs, crossing of national borders, boarding of planes among others. In the absence of robust schemes, these systems are vulnerable to the wiles of an impostor. Current systems are based on the three vertex of the authentication triangle which are, possession of the token, knowledge of a secret and possessing the required biometric. Due to weaknesses of the de facto password scheme, inclusion of its inherent keystroke rhythms, have been proposed and systems that implement such security measures are also on the market. This correspondence investigates possibility and ways for optimising performance of hardened password mechanism using the widely accepted Neural Network classifier. It represents continuation of a previous work in that direction.

LOTKIP: Low Overhead TKIP Optimization for Ad Hoc Wireless Networks

Temporal Key Integrity Protocol (TKIP) is a provisional solution for Wired Equivalent Privacy (WEP) security loopholes present in already widely deployed legacy 802.11 wireless devices. In this work, we model and analyse the computational complexity of TKIP security mechanism and propose an optimised implementation, called LOTKIP, to decrease processing overhead for better energy efficient security performance. The LOTKIP improvements are based on minimising key mixing redundancy and a novel frame encapsulation with low overhead. We simulate and compare LOTKIP with baseline TKIP in terms of complexity and energy consumption for ad hoc wireless network security. From simulation results, we demonstrate that LOTKIP executes with lower computational complexity, hence, with faster encryption time and more energy-efficient.